def trainPredModel(data):
'''
Split dataset to training and evaluation (20% evaluation split)
Fit model to training data, and generate a doc with evaluation
as well as the summary of errors
'''
process = ProcessData(sentdata)
padded_sequence, tokenizer, labels = process.createTrainData()
ft = createFTEmbedding()
ft.processDict(tokenizer)
x_train, x_val, y_train, y_val = train_test_split(padded_sequence, labels)
model = compileModel(padded_sequence, labels, tokenizer,
ft.embedding_matrix)
model.fit(x_train,
y_train,
batch_size=20,
epochs=10,
validation_data=(x_val, y_val))
y_pred = model.predict(x_val)
translateEvalData(y_pred, y_val, x_val, tokenizer)
return model, x_train, x_val, y_train, y_val, y_pred
